[dm-devel] dm-crypt parallelization patches

[Mikulas Patocka]

On Tue, 9 Apr 2013, Vivek Goyal wrote:

> On Tue, Apr 09, 2013 at 04:32:28PM -0400, Mikulas Patocka wrote:
> 
> [..]
> > Generally, we shouldn't associate bios with "current" task in device 
> > mapper targets. For example suppose that we have two stacked dm-crypt 
> > targets:
> >
> > In the "current" process pointer in lower dm-crypt target's request 
> > function always points to the workqueue of the upper dm-crypt target that 
> > submits the bios. So if we associate the bio with "current" in the lower 
> > target, we are associating it with a preallocated workqueue and we already 
> > lost the information who submitted it.
> > 
> > You should associate a bio with a task when you create the bio and "md" 
> > and "dm" midlayers should just forward this association to lower layer 
> > bios.
> 
> bio_associate_current() return -EBUSY if bio has already been associated
> with an io context.
> 
> So in a stack if every driver calls bio_associate_current(), upon bio
> submission, it will automatically amke sure bio gets associated with
> submitter task in top level device and calls by lower level device
> will be ignored.

The stacking drivers do not pass the same bio to each other.

The stacking driver receives a bio, allocates zero or more new bios and 
sends these new bios to the lower layer. So you need to propagate 
ownership from the received bio to newly allocated bios, you don't want to 
associate newly allocated bio with "current" process.

> Lower level devices I think just need to make sure this context
> info is propogated to cloned bios.
> 
> 
> [..]
> > +#ifdef CONFIG_BLK_CGROUP
> > +	tio->clone.bi_ioc = ci->bio->bi_ioc;
> > +	tio->clone.bi_css = ci->bio->bi_css;
> 
> You also need to take references to ioc and css objects. I guess a helper
> function will be better. May be something like.

The lifetime of the "tio" structure is shorter than the lifetime of 
"ci->bio". So we don't need to increment reference.

We only need to increment reference if we copy ownership to a new bio that 
has could have longer lifetime than the original bio. But this situation 
is very rare - in most stacking drivers the newly allocated bio has 
shorter lifetime than the original one.

> bio_associate_bio_context(bio1, bio2)
> 
> And this initialize bio2's context with bio1's context.

Yes, that would be ok.

> Thanks
> Vivek

Mikulas